Multiobjective Optimization of the Breathing System of an Aircraft Two Stroke Supercharged Diesel Engine

Available online at www.sciencedirect.com ScienceDirect

Energy Procedia 82 ( 2015 ) 31 – 37

ATI 2015 - 70th Conference of the ATI Engineering Association Multiobjective optimization of the breathing system of an aircraft two stroke supercharged Diesel engine

Antonio Paolo Carluccia,*, Antonio Ficarellaa, Domenico Laforgiaa, Gianluca Trulloa

aDepartment of Engineering for Innovation, University of Salento, Via per Monteroni, Lecce 73100, Italy

Abstract

One of the factors limiting the utilization of piston internal combustion engines for aircraft propulsion is the performance decrease increasing the altitude of operation. This is due to the negative effect of air density reduction increasing the altitude on cylinder filling. A solution to this problem is represented by the engine supercharging. Unfortunately, in two stroke engines, the cylinder filling efficiency is antithetical to the cylinder scavenging efficiency. With the aim of guaranteeing an optimal balance between engine performance and specific consumption, an engine breathing system optimization is needed. In this work, the results obtained running a multi-objective optimization procedure aiming at performance increase and fuel consumption reduction of an aircraft two stroke supercharged diesel engine at various altitudes are analyzed. During the optimization procedure, several geometric parameters of the intake and exhaust systems as well as geometric and operating engine parameters have been varied. Then, a multi-objective optimization algorithm based on genetic algorithms has been run to obtain the configurations optimizing the engine performance at Sea Level (take-off conditions) and fuel consumption at 10680 m (cruise conditions).

© 20152015 The The Authors. Authors. Published Published by Elsevier by Elsevier Ltd. This Ltd. is an open access article under the CC BY-NC-ND license (Selectionhttp://creativecommons.org/licenses/by-nc-nd/4.0/ and/or peer-review under responsibility). of ATI Peer -review under responsibility of the Scientific Committee of ATI 2015 Keywords: 2-strokes Diesel engine; scavenging, multi-objective optimization

1. Introduction

In two stroke engines, the cylinder filling and emptying phases are driven by the difference between inlet and outlet pressure. Therefore, the supercharging system of a two stroke engine determines both the air density increase and the cylinder scavenging. Due to the strict interaction among supercharging,

* Corresponding author. Tel.: -39 0832 297751; fax: +39 0832 297777. E-mail address: [email protected].

1876-6102 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Scientific Committee of ATI 2015 doi: 10.1016/j.egypro.2015.11.879 32 Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37

scavenging and combustion, two stroke engines are very sensitive to the variation of engine working and geometric parameters as well as ambient conditions with altitude 0, 0. In order to obtain the best cylinders breathing mechanism, different supercharging architectures for a two stroke engine prototype (a single mechanical compressor supercharger; a turbocharger with a crankcase scavenging pump; a mechanical compressor combined with a turbocharger) have been numerically modeled and their performance compare dat sea level in 0. Performance are optimized using a turbocharger upstream a second compression stage, due to both compression power reduction and trapping efficiency increase. Multi-objective optimization processes based on genetic algorithms are often used to improve engine performance and to study the effect of input parameters variation on the engine behavior 0. In this work, the results obtained running a multi-objective optimization procedure aiming at reducing fuel consumption and increasing performance of an aircraft two stroke supercharged diesel engine at difference altitudes are analyzed.

2. Engine Model and Optimization Procedure

The engine analyzed in the present work is a two stroke Diesel Engine – composed by six cylinders arranged in boxer configuration in two independent banks – for aircraft propulsion. Each bank is fed by a supercharging system composed by two turbochargers, an after-cooler and a common intake plenum. The scavenging system is “Uniflow”, with 14 inlet ports and 2 exhaust valves per cylinder. The engine main specifications are reported in Table 1, while its scheme is reported in Errore. L'origine riferimento non è stata trovata..The 0D-1D model of the engine has been realized in AVL BOOST software v2011.2. More details about the software are reported in 0 while details about the engine model and calibration can be found in 0.

Table 1. Engine main specifications

Cycle Two-Stroke Diesel Uniflow Bore/Stroke ratio 1 Compressor Ratio 17.2:1 Injection System Common Rail Design Engine Speed 2000 rpm

Fig. 1. Engine model (single bank) Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37 33

With the aim of increasing the engine output power and reducing fuel consumption redesigning the intake and exhaust systems, the engine model was run varying several parameters on a range of allowed values. In Table 2 the engine parameters varied during simulations are listed. These parameters have been varied running the model at two different altitudes: Sea Level (SL) and 10670 m. Concerning the exhaust valves, once fixed the values for Exhaust Valve Opening (EVO), Closing (EVC) and maximum lift (HEV), the lift variation with Degrees Crank Angle (DCA) has been calculated through a routine implemented in Matlab and then transferred to the engine model in BOOST. Due to the considerably high number of combinations of parameters, an approach based on genetic algorithms was chosen in order to find the combinations leading to a better engine performance. Therefore, the engine thermo-dynamic model was interfaced with ESTECO ModeFrontier software 0. During the optimization procedure, several constraints were also introduced in order to discard those solutions not technically feasible: maximum in-cylinder pressure not higher than150 bar; maximum temperature at the after cooler outlet not higher than 473.15 K; maximum exchanged heat power in the after cooler not higher than 40 kW.

Table 2. Engine input parameters

Actual Lower Upper Central Input Parameter Unit Acronym Step Value Bound Bound Value Volumetric Compression Ratio [-] CR 17.2 13 18 15.5 0.2 Exhaust Valve Opening (after TDC) [DCA] EVO 80 60 110 85 5 Exhaust Valve Closing (after TDC) [DCA] EVC 250 220 270 245 5 Exhaust Valve Maximum Lift [mm] HEV 12 8 14 11 0,5 Intake Port Opening (after TDC) [DCA] IPO 115 100 130 115 2.5 Port Section [mm2] A 265 200 300 250 5 Number of Inlet Ports [-] n 14 10 19 14.5 1 Distance of Inlet Port from BDC [mm] HBI 2.5 1 3 2 0.25 Air Fuel Ratio [-] AFR 20 17 27 22 0.5 Pressure Ratio – High Pressure [-] PRHP 1.7 1.2 2.2 1.7 0.05 Compressor @ SL Pressure Ratio – Low Pressure [-] PRLP 1.75 1.2 2.2 1.7 0.05 Compressor @ SL Pressure Ratio – High Pressure [-] PRHP 2.3 2.2 3.4 2.8 0.05 Compressor @ 10670 m Pressure Ratio – Low Pressure [-] PRLP 3 2.2 3.4 2.8 0.05 Compressor @ 10670 m

As previously said, the output variables to be optimized were the engine output power and specific fuel consumption. The Design of Experiment generated 30 random starting solutions; for each one of them, the genetic algorithm generated 70 combinations. The resulting 2100 configurations defined the Pareto Front, where every solution represents a non-dominated relative optimum 0.Once the Pareto front had been created, the best solutions have been defined as described in the following.

3. Data Analysis

The Pareto front represents 0 a series of optimum points defined as “non-dominated solutions”. This means that each point composing the Pareto front is the best achievable solution using a specific set of 34 Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37

input parameters. However, given the significant amount of input parameters and levels set for the optimization procedure, a statistical approach was chosen to interpret the obtained results. In order to manage the high number of input data and their non-normal distribution trend, the t-Student distribution was chosen, and two meaningful statistic parameters have been calculated: the Effects Size (ES) and the Significance (S). As suggested by Ferguson 0, ES is an estimate of the magnitude of the effects of the association between two or more variables. The computation of this parameter is reported by Cohen 00;in this work it can be assumed, for sake of simplicity, that the magnitude of the effects that an input parameter has on an output one is directly proportional to the ES module. More specifically, when ES is positive, a direct relation between the input and the output occurs, whereas when ES is negative the relation is inverse. The latter statistic parameter is S. The significance level of a statistical hypothesis test is a fixed probability (referred to a fixed threshold value set as 0.05) of rejecting the null hypothesis 0 which represents a statistic value that has to be accepted or rejected according to the computed S value. To prevent from false claims the significance value has to be as low as possible and eventually equal to 0.

4. Discussion of Results

As previously said, ES and S can help in recognizing which inputs have the largest effects on engine power (Pb) and specific fuel consumption (BSFC). In Table 3 all inputs and corresponding values of ES and S are listed for flight levels equal to 0 m (SL) and 10680 m.

Table 3. Statistic parameters ES and S associated to each input – SL and 10680 m cases

Statistic Input Parameters Goal Function Parameter CR AFR EVC EVO HEV IPO A n HBI PRLP PRHP Effect Size 11.6 -53.6 16.39 3.32 -2.28 -1.53 1.9 -7.5 -24.9 32.26 49.76 Pb @SL Significance 0 0 0 0 0.04 0 0.052 0 0 0 0 Effect Size -17.7 29.5 -11.3 -18 4.98 0.49 1.69 7.74 14.4 -18.8 -28.2 BSFC @SL Significance 0 0 0 0 0 0.295 0.017 0 0 0 0 Effect Size 11.83 -34.7 19.91 -4.92 0.14 12.16 -19.3 8.41 11,74 26.04 28.5 Pb @10760 m Significance 0 0 0 0 0.43 0 0 0 0 0 0 Effect Size -18.7 35.15 -15.4 -13.8 0.88 -16.8 19.84 -12.7 - 9,36 -31.6 -34.1 BSFC @10760 m Significance 0 0 0 0 0.22 0 0 0 0 0 0

Specifically, considering Pb at SL, the relationship between CR, AFR, EVC, n, HBI, PRHP, PRLP and the goal function is strong and this is proved by the relatively high ES values. In cruise condition, on the other hand, the ES related to IPO and A increase considerably and, furthermore, all the ES values of the breathing system parameters changed their sign (i.e. IPO, n, HBI and A).Considering the BSFC minimization process at SL, instead, two aspects can be immediately observed: the first one is that all the parameters reaching the highest ES values maximizing Pb changed their sign; the latter is that EVO became more influent on BSFC than on Pb. In this case too, the magnitude of the HBI and A ES grows considerably and all the parameters related to the intake system changed their sign in cruise conditions.

Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37 35

4.1. Multi Criteria Decision Making

As previously said, once the optimization process has been run, a high number of solutions compose the Pareto front. It is necessary to choose, among all these values, the ones defining the final breathing system configuration. This is not an immediate decision because, lying on the Pareto front, every solution is not dominated by any other. A possible way to reach this goal is to set several design preferences on the variation range assigned to each parameter, or a higher weight to some of the output optimized values. In the case under analysis, referring to the target values, the most relevant parameter at SL and take-off conditions, is the output power, whereas the BSFC is the most relevant one at 10680 min cruise conditions. Consequently, among all the solutions at SL, the one with the highest output power has been chosen, while, in cruise conditions, the one with the lowest BSFC has been selected. Both the solutions and relative input values are reported in Table 4. In particular, engine brake power Pb is reported in % of Pb target either at sea level and at 10680 m; the engine BSFC are also reported in % of a target value calculated as average of the target values at the two analyzed altitudes. Beside the tables reporting the solutions, the two Pareto fronts for SL and 10680 m condition in which the solutions listed before have been highlighted are shown in figure 2.

Table 4. Flight Level 0 m and 10680 m: maximum engine Pb(case A) and minimum BSFC(case B)

Flight Pb BSFC Trapping Scavenging A AFR CR EVC EVO HBI IPO HEV PRLP PRHP n level [%] [%] Efficiency Efficiency 0 m 142.34 99.75 240 17 16 270 110 1 115 10.5 1.7 2.1 10 0.92 0.69 10680 m 120.77 92.92 225 18.5 18 240 95 2 122.5 11 2.7 2.7 17 0.89 0.75

170 160

150 A 140 (a)(a) (b) 130 B 120 [%] [%] b 110 b 100 90 Normalized P Normalized Normalized P Normalized 80 70

60 50

30 40 50 60 70 80 90 100 110 120 130 140 150 50 60 70 80 90 100 110 120 130 140 150

Normalized BSFC [%] Normalized BSFC [%]

Fig. 2. Pareto Front for Flight Level Equal to 0 m (a) and 10680 m (b); the highlighted points represent Pbcase at 0 m (A) and minimum BSFC case at 10680 m (B)

From values reported in Table 4, it can be deduced that the two turbochargers compression ratios must have a very close value, with PRLP slightly lower than PRHP (at 0 m FL) or coincident (at 10680 m FL). This condition was somehow expected since it was shown to be related to the engine operation characterized by the lowest air compression work 0. Concerning the engine breathing system geometry, based on the previous statistical analysis, it was shown that the most significant intake and exhaust parameters are: the number of inlet ports, the single port section and the distance between the port and the piston at BDC (respectively n, A and HBI) for the inlet; the opening and closing angles(EVO and EVC) for the exhaust. The configuration maximizing the output power in take-off conditions (see Table 4) requires a number of ports as low as possible (n = 10) while the single port section is set at an average 36 Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37

value (A = 240 mm2). For the exhaust system, instead, EVO and EVC must be respectively equal to 110 and 270 DCA. The configuration minimizing the BSFC in cruise condition is characterized by different values for the previous parameters; in particular, the number of inlet ports must be higher (n= 17) while the section of the single port assumes a much lower value (A = 225 mm2). Also EVO and EVC shifted towards more advanced angles (95 and 240 DCA).The difference between the two sets of input scan be motivated by the different output to be optimized together with the different operating conditions. The intake parameters oriented towards power maximization aim to obtain a higher quantity of fresh charge trapped into the cylinder; in fact, using large sections and low number of ports,a low intake air speed and a better scavenging process is obtained. In cruise conditions, pursuing BSFC minimization, the intake parameters – high inlet port number and small single port section - force the intake air into the cylinder with a higher speed to generate a better trapping efficiency. Similarly, the exhaust system parameters in the case of power maximization is characterized by valve opening and closing angles higher than in cruise conditions. Moreover, the valves opening period is longer at SL than at 10680 m. Maximizing the output power requires to use the highest portion of the expansion stroke and, for this reason, the exhaust valves must be opened later than in the minimum consumption case. Furthermore, the valves opening period must be longer because a longer time is required to effectively scavenge the cylinder with the slower intake air due to the larger inlet ports section and the higher exhaust pressure at EVO. On the other hand, longer valves opening period is reflected in lower exhaust temperature implying lower thermal energy available for the turbines and compressors and consequently lower values of efficiency.

5. Conclusions

This work describes the optimization process and the results obtained for a breathing system of an aircraft two stroke supercharged Diesel engine. Aim of the optimization process was to increase the engine output power and reduce the fuel consumption. Several input parameters, supposingly influent on the goal functions, have been varied on an allowed range of values. The engine has been modeled and simulated in take-off and cruise conditions varying the input parameters based on a genetic algorithms optimization procedure. Based on the solutions found by this procedure, it was possible to analyze the influence of each parameter on the goal functions. Afterwards, the configurations leading to the highest power in take-off condition and to the lowest fuel consumption in cruise conditions have been further analyzed. As a consequence of this analysis, it has been noticed that, even if with opposite influence relation, the input parameters which influence the goal functions are the same. It is necessary to emphasize that the intake system parameters become more significant in cruise than in take-off, because the decrease in inlet air density requires appropriate input parameters that guarantee a good compromise between scavenging and trapping processes efficiency.

References

[1] Ravi M, Marathe A. Effect of Port Sizes and Timings on the Scavenging Characteristics of a Uniflow Scavenged Engine. SAE Paper 920782; 1992. [2] Carlucci AP, Ficarella A, Laforgia D, Renna A. Supercharging system behavior for high altitude operation of an aircraft 2- stroke Diesel engine.Energy Conversion and Management; 2015;101:p. 470-480. [3] Knoll R. AVL Two-stroke Diesel Engine. SAE Paper 981038; 1998 Antonio Paolo Carlucci et al. / Energy Procedia 82 ( 2015 ) 31 – 37 37

[4] Carlucci AP, De Risi A, Donateo T, Ficarella A. A Combined Optimization Method for Common Rail Diesel Engines.Proceedings of 2002 Spring Technical Conference of the ASME Internal Combustion Engine Division, Rockford (Illinois), April 14-17, 2002. [5] AVL BOOST. Users Guide for AVL BOOST. AVL List GmbH, 2008. [6] ESTECO.ModeFrontierUser’s Manual. [7] Ferguson CJ. An Effect Size Primer: A Guide for Clinicians and Researchers. Professional Psychology: Research and Practice, Oct 2009; Vol 40(5): p. 532-538. [8] Cohen J. Statistical power analysis. Current Directions in Psychological Science; Jun, 1992; Vol. 1, No. 3: p. 98-101 [9] Cohen J. A power primer. Psychological Bulletin; Jul. 1992; Vol 112(1), p. 155-159. [10] Frick RW. The appropriate use of null hypothesis testing. Psychological Methods; 1996; p. 379-390.

Biography Paolo Carlucci graduated in 2000 with Honors and received the Ph.D. degree in “Energy Systems and Environment” in 2004. In November 2005 he joined the Engineering Faculty at University of Lecce as Assistant Professor. His research interests are combustion and emissions in diesel and dual-fuel internal combustion engines.